Select the method or formula of your choice.

The calculation of sample size, n, and critical distance, k, depends on the number of specification limits given and whether standard deviation is known.

The sample size is given by:

The critical distance is given by:

where:

Term | Description |
---|---|

Z_{1} | the (1 – p_{1}) * 100 percentile of the standard normal distribution |

p_{1} | the Acceptable Quality Level (AQL) |

Z_{2} | the (1 – p_{2}) * 100 percentile of the standard normal distribution |

p_{2} | the Rejectable Quality Level (RQL) |

Z_{α} | the (1 – α) * 100 percentile of the standard normal distribution |

α | the consumer's risk |

Z_{β} | the (1 – β ) * 100 percentile of the standard normal distribution |

β | the producer's risk |

Notation is the same as for the section for a single specification limit and a known standard deviation. The sample size is given by:

The critical distance is given by:

Notation not defined below is the same as for the case for a single specification limit and a known standard deviation. First Minitab calculates z:

Then Minitab finds p* from the standard normal distribution as the upper tail area corresponding to z. This is the minimum probability of defective outside one of the specification limits.

The method that Minitab uses for the calculation of the sample size and the critical distance depends on this value of p*.

Let p_{1} = AQL, p_{2} = RQL

- If 2p* ≤ (p
_{1}/ 2), then the two specifications are relatively far apart and calculations follow the single limit plans. - If p
_{1}/ 2 < 2p* ≤ p_{1}, then the two specifications are not relatively far apart, but are still not so close that the minimum probability of defective can be found for certain mean values. Minitab performs an iteration to find sample size and critical distance.

Let

μ = μ_{0}+ m * h, where h = σ/100

Let m = 1, 2, ...300. For each μ calculate:

where Φ is the cumulative distribution function of the standard normal distribution. If Prob (X<L) + Prob (X>U) is extremely close to p_{1}, then Minitab uses the larger value between Prob (X<L) and Prob (X>U) to find sample size and accept number.

Suppose Prob (X<L) is the larger value, let p_{L} = Prob (X<L).

The sample size is given by:

The critical distance is given by:

where:

Z_{pL} = the (1 – p_{L}) * 100 percentile of the standard normal distribution.

If we already use all the m values but the corresponding probabilities do not contain p_{1}, then p_{1} is too large, which means that the mean of the measurements is far away from the midpoint of the interval [L, U]. In this case, we can use a method for a single specification limit and Z_{pL} = Z_{1}. Z_{1} has the same definition as for the single specification limit case.

- If p
_{1 }< 2p* < p_{2}, then the specifications of the plan must be reconsidered because the minimum probability defective determined by the two specification limits and the standard deviations is larger than the acceptable quality level p_{1}. You can reject the lot or consider a plan with a slightly larger probability of defective than p_{1}. - If 2p* ≥ p
_{2}, then the lot should be rejected; the minimum probability of defective determined by the two specification limits and the standard deviation is larger than the rejectable quality level. You can reject the lot without testing any products.

Term | Description |
---|---|

L | the lower specification limit |

U | the upper specification limit |

σ | the known standard deviation |

Notation is the same as for the previous sections. Minitab lets the critical distance be the value as given in the case of two separate single-limit plans:

The sample size is given by:

If *n* > 2, then Minitab calculates the MSD with the following steps^{1}.

- Let
- Then,
where Beta is the cumulative distribution function of a beta distribution with shape parameters
*a*and*b*. Here, . - Define
- Then,
where Beta
^{-1}is the inverse cumulative distribution function of the beta distribution from step 2.

If *n* ≤ 2, The Maximum Standard Deviation (MSD) is incalculable.

Let p be the probability of defective which is the x value of a point on an OC curve.

- Single lower specification limit and known standard deviation
- Prob (X < L) = p.
- Single upper specification limit and known standard deviation
- Prob (X > L) = p.

First Minitab calculates z

Then finds p* from the standard normal distribution as the upper tail area corresponding to z. This is the minimum probability of defective outside one of the specification limits.

The method Minitab uses for the probability of acceptance depends on this value of p*.

Let p_{1} = AQL, p_{2} = RQL

- If 2p* ≤ (p
_{1}/ 2), then the two specifications are relatively far apart and calculations for sample size and critical distance follow the single limit plans. - If p
_{1}/ 2 < 2p* ≤ p_{1}, then the two specifications are not relatively far apart, but are still not so close that the minimum probability of defective can be found for certain mean values.

For any given p, Minitab finds the mean, μ, of the measurements using a grid search algorithm. Then,

When you have both upper and lower specification limits, but do not know the standard deviation, Minitab uses the OC curve for the single-limit plan to approximate the double specification limits case. The OC curve derived for a single-limit plan with specified p_{1}, p_{2}, α, and β is the lower limit of the band of OC curves for a two-sided specification plan with the same p_{1}, p_{2}, α, and β and for most practical cases can be taken as the OC curve for the two-sided plan. See Duncan^{1}.

- Duncan (1986). Quality Control and Industrial Statistics, 5
^{th}edition.

Term | Description |
---|---|

n | sample size |

k | critical distance |

σ | known standard deviation |

Z_{p} | the (1 - p)^{th} percentile from the standard normal distribution |

Φ | the cumulative distribution function of the standard normal distribution |

T |
is non-central t distributed with degrees of freedom = n – 1, and the non central parameter, |

L | lower specification limit |

U | upper specification limit |

The probability of rejecting (P_{r}) describes the chance of rejecting a particular lot based on a specific sampling plan and incoming proportion defective. It is simply 1 minus the probability of acceptance.

P_{r} = 1 – P_{a}

where:

P_{a} = probability of acceptance

The average outgoing quality represents the quality level of the product after inspection. The average outgoing quality varies as the incoming fraction defective varies.

Term | Description |
---|---|

P_{a} | probability of acceptance |

p | incoming fraction defective |

N | lot size |

n | sample size |

The average total inspection represents the average number of units that will be inspected for a particular incoming quality level and probability of acceptance.

Term | Description |
---|---|

P_{a} | probability of acceptance |

N | lot size |

n | sample size |

The acceptance region is calculated only when both specifications are given and the standard deviation is unknown. Go to the section on sample size and critical distance to find definitions for *n* and *k*, respectively, and to see the notation for the equations..

On the acceptance region plot, the x-axis is the sample mean and the y-axis is the sample standard deviation. The acceptance region is formed by 3 functions of the sample standard deviation and the sample mean in addition to the Maximum Standard Deviation (MSD). For any values of the sample mean where the values of the sample standard deviation exceeds the MSD, the upper bound of the acceptance region is the MSD.

For cases that are near the upper or lower specification limits, the acceptance region is bounded by these two functions:

As the values of the sample mean get closer to the middle of the specification limits, the coordinates of the upper boundary of the acceptance region are calculated with the following steps:

- Let .
- Then where Beta is the cumulative distribution function of a beta distribution with shape parameters
*a*and*b*. Here, .. - Define pairs of proportions p
_{01}and p_{02}, that satisfy p_{02}+ p_{01}= p* - Next,
where Beta
^{-1}is the inverse cumulative distribution function of the beta distribution from step 2. - For , the mean and standard deviation coordinates are the following equations: